AI Systems Amplify Age and Gender Bias Against Women Online
A Stanford-led study of 1.4 million images from Google, Wikipedia, IMDb, Flickr and YouTube, plus nine large language models including ChatGPT, found that women professionals are consistently depicted as younger and less experienced than men, particularly in high-status occupations. The research shows AI algorithms amplify this bias when generating and evaluating job candidates.
What happened
A study led by Douglas Guilbeault at Stanford University analyzed 1.4 million images and videos from Google, Wikipedia, IMDb, Flickr and YouTube across 3,495 occupational categories, plus nine large language models including ChatGPT. The analysis found women are consistently depicted as significantly younger than men, with larger disparities in high-status, high-income professions.
When ChatGPT evaluated approximately 40,000 resumes, it assumed women were younger and less experienced, while rating older male applicants as more qualified. Two experiments with roughly 500 participants confirmed that exposure to these biased online images influences social perceptions.
Reported impact
The bias perpetuates workplace stereotypes that women in professional roles are less experienced than their male counterparts, despite U.S. census data showing virtually identical average working ages for both sexes. This distortion is particularly harmful in high-status professions where age and appearance carry greater weight in hiring and promotion decisions.
Index analysis
The research reveals that digital platforms and AI models trained on internet data inherit and amplify pre-existing gender and age biases. According to the researchers, social beauty standards disproportionately pressure women to appear young, while age signals status, authority and power—qualities culturally associated with men. Structural factors including the 'glass ceiling' also contribute, as young women are hired but promoted less frequently than young men, concentrating power and leadership among older men.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.